Yi 34B

🚀 Yi - Building the Next Generation of Open - Source and Bilingual LLMs

The Yi series is a set of open - source large language models developed from scratch. Trained on a 3T multilingual corpus, it shows excellent performance in language understanding, commonsense reasoning, and reading comprehension, ranking highly in multiple benchmarks.

🚀 Quick Start

Choose your path

You can start using Yi through different methods such as pip, docker, llama.cpp, conda - lock, or the web demo.

Quick start - pip

Quick start - docker

Quick start - llama.cpp

Quick start - conda - lock

Web demo

✨ Features

• High - performance in benchmarks: The Yi - 34B - Chat model ranked second on the AlpacaEval Leaderboard (as of January 2024), and the Yi - 34B model ranked first among open - source models in both English and Chinese on various benchmarks (as of November 2023).
• Continuous improvement: The long - text capability of Yi - 34B - 200K has been enhanced, and new models like Yi - 1.5 series, Yi - 9B, etc., have been continuously open - sourced.
• Multiple model types: There are chat models and base models to meet different application scenarios.

📦 Installation

The installation methods include using pip, docker, llama.cpp, and conda - lock. You can choose the appropriate method according to your needs.

💻 Usage Examples

Basic Usage

You can try the chat models interactively on platforms like Hugging Face and Replicate. For example, when using the Yi - 34B - Chat model:

Input: "hi"
Output: " Hello! How can I assist you today?"

Advanced Usage

The base models can be fine - tuned to meet specific requirements. You need to ensure that you meet the software and hardware requirements.

📚 Documentation

What is Yi?

Introduction

• The Yi series models are the next - generation open - source large language models trained from scratch by 01.AI.
• They are targeted as bilingual language models and trained on a 3T multilingual corpus, showing excellent performance in multiple aspects.
• Although based on the Transformer architecture like Llama, the Yi series models are not derivatives of Llama as they do not use Llama's weights.

Models

Chat models

Base models

The base models include Yi - 6B, Yi - 34B, Yi - 6B - 200K, Yi - 34B - 200K, etc.

Model info

The models are trained on a 3T multilingual corpus and are based on the Transformer architecture.

News

• 2024 - 07 - 29: The Yi Cookbook 1.0 is released, featuring tutorials and examples in both Chinese and English.
• 2024 - 05 - 13: The Yi - 1.5 series models are open - sourced.
• 2024 - 03 - 16: The Yi - 9B - 200K is open - sourced.
• 2024 - 03 - 08: Yi Tech Report is published.
• 2024 - 03 - 07: The long - text capability of the Yi - 34B - 200K has been enhanced.
• 2024 - 03 - 06: The Yi - 9B is open - sourced.
• 2024 - 01 - 23: The Yi - VL models, [Yi - VL - 34B](https://huggingface.co/01-ai/Yi-VL-34B) and [Yi - VL - 6B](https://huggingface.co/01-ai/Yi-VL-6B), are open - sourced.
• 2023 - 11 - 23: Chat models are open - sourced, and the Yi Series Models Community License Agreement is updated to v2.1.
• 2023 - 11 - 08: Invited test of Yi - 34B chat model.
• 2023 - 11 - 05: The base models Yi - 6B - 200K and Yi - 34B - 200K are open - sourced.
• 2023 - 11 - 02: The base models Yi - 6B and Yi - 34B are open - sourced.

How to use Yi?

Fine - tuning

The base models can be fine - tuned to adapt to specific tasks.

Quantization

There are quantization methods such as 4 - bit and 8 - bit quantization for the models.

Deployment

You need to meet the software and hardware requirements for deployment.

FAQ

The frequently - asked questions can help you solve common problems.

Learning hub

You can grow at [Yi Learning Hub](#learning - hub).

Why Yi?

Ecosystem

• Upstream: Based on the Transformer and Llama architectures, it benefits from the open - source community.
• Downstream: There are serving, quantization, fine - tuning, and API services.

Benchmarks

• Base model performance: The Yi - 34B model shows excellent performance on various benchmarks.
• Chat model performance: The Yi - 34B - Chat model ranks highly on the AlpacaEval Leaderboard.

Tech report

You can check out the Yi Tech Report.

Citation

If you use Yi in your research, please cite the relevant report.

Who can use Yi?

Yi can be used by researchers, developers, and other users who need language - related services.

Misc.

Acknowledgments

Thanks to the Transformer and Llama open - source communities.

Disclaimer

Please refer to the relevant license agreement for details.

License

The license is Apache - 2.0.